FYI: latest patches to processonelog

[Mark Linimon]

On pointyhat-west, we've changed the default build environment to
jails that are restricted from talking to the Internet after the
fetch phase.  (If we allow this, who knows what the software could
actually be doing?)  Of course, as one might expect, this triggered
some port failures.  These ports were all Doing The Wrong Thing IMVHO.

In any case, I now have a patch to processonelog in the pointyhat
codebase that catches these cases and labels them 'fetch' instead
of '???'.

Finally, as a backstop, I added a line to catch "package build
never completed" and flag it as another case of 'cluster'.  Since
these are rare (and probably only seen on the pointyhat codebase),
the test is at the bottom for performance reasons.

Here is the list of new failures:

  http://wiki.freebsd.org/PortsFailingInJails

and here is the code change (spans 2 revisions):

  http://svnweb.freebsd.org/base/projects/portbuild/scripts/processonelog?r1=227988&r2=236031

I don't know if tinderbox has picked up the new jail code yet, but in
any case, this change should probably be merged into the tinderbox
equivalent, to try to keep things from drifting.

Thanks.

mcl

* note: if you fix one of the failing ports, please move it to the
"already fixed" section, since that will save me time when scanning
complete sets of errorlogs.  Thanks.